OpenMx 2.17 released!

We are pleased to announce the official release of OpenMx version 2.17.2. Click here for instructions on how to install the package from our repository. As usual, our repository has package binaries for Windows and MacOS, and source tarballs for Linux/GNU and other non-Mac Unix-likes, all of which come with the proprietary NPSOL optimizer. Alternately, users may install the fully open-source build of the new version from CRAN.

New Features and Performance Improvements Since v2.15:

• When the input data are of type="cor", if (and only if) the MxModel is of type="RAM" or type="LISREL", OpenMx now automatically adds constraints to keep the diagonals of the model-expected correlation matrix equal to 1.0. This ensures that the number of observed statistics is correct for correlation-matrix data with those model types.
• The mxPearsonSel* functions now have two interfaces for specifying which dimensions on which to condition. Under the old interface, newCov is a submatrix of origCov, and the dimnames are matched to determine which partition of origCov to replace with newCov. Under the new interface, newCov is the same dimension as origCov. The matrix entries are inspected to determine which entries have changed. The changed entries determine which partition of origCov to replace with newCov.
• mxData() has a new argument, naAction, which governs how missing values in raw datasets are handled.
• For models of type="LISREL", if argument manifestVars to mxModel() is a character vector instead of a list, then it is now treated as the names of the model's endogenous manifest variables.
• There is now a logLik() S3 method for MxModels that use the WLS fitfunction; it returns the model chi-square statistic.
• There is now an AIC() S3 method for MxModels that use the WLS fitfunction; it returns the model chi-square statistic plus k times the chi-squared degrees of freedom, where k defaults to 2.
• The man page example of omxSelectRowsAndCols() has been clarified.

Bug-fixes and tweaks Since v2.15:

• In the "free parameters" table of summary() output, OpenMx formerly would give the row and column names, if there were any, when identifying the MxMatrix element in which a free parameter first appeared in the model. At some unknown point, this behavior was lost, and summary() output would always use the row and column numbers, even if names were available. The old behavior has now been restored.
• Dimnames of MxMatrices and MxAlgebras are now copied through elementwise algebra operations.
• OpenMx now detects requests for confidence intervals on objects that do not exist in the MxModel, and gives a specific error message.

Known issues

• summary() still counts redundant equality MxConstraints as though they were nonredundant (which is not a new issue, and likely has always existed in OpenMx). Therefore, it can get the model degrees-of-freedom wrong if redundant equality constraints are present in the model.